Gear Up On the Runway Causes 1,200-Metre Skid
Lennart Meri Tallinn Airport (Tallinna lennujaam) is the largest airport in Estonia. It serves as a hub for Estonian airline Nordica as well as the secondary hub for AirBaltic and LOT Polish Airlines.
Over the past year, Tallinn has worked to expand its single asphalt-concrete runway to 3,070 metres but at the time, in 2010, runway 08 had a length of 2,800 metres (9,186 feet).
The take-off distance of the Antonov An-26 is around 1,250 metres (4,000 feet), less than half of the available runway at Tallinn at that time.
And still, somehow, it wasn’t enough.
Flight 3788 was a scheduled cargo flight operated by Exin Co. Ltd., a Polish charter airline. Exin was established in 1991 and currently operates two aircraft, both An-26s.
This is a photograph of SP-FDP taken by Mikko Mertanen in Helsinki just three months before the incident.
SP-FDP was an Antonov An-26B, a twin-engine transport plane with two turboprop engines and one auxiliary turbojet engine. The An-26 was designed and produced in the Soviet Union from 1969 to 1985 and was very popular as a cargo aircraft because of the retractable cargo ramp. The An-26B is the civilian cargo version of the aircraft equipped with roller gangs which can be swung up against the cabin walls when not in use. The two turbo-prop power plants deliver higher thrust than the standard model and an additional small turbojet adds thrust to act as a climb rate/high altitude cruise booster.
On the 25th of August 2010, the aircraft was scheduled for a cargo flight from Tallinn to Helsinki.
There were four crew on the flight deck: the captain, the first officer, the navigator and the engineer. The crew had been put together five days earlier. The flight engineer was new to the company and, before joining it, he’d had a three month gap in his flight activities.
All crew had passed the Crew Resource Management training given by the company in May-June 2010; that is, all of them had covered CRM training within six months of the accident.
The crew was based in Helsinki, where Exin had a guest house. Rest and duty time rules were observed. None of the crew members showed any signs of alcohol or medical issues which might have inhibited their ability to act as crew.
There’s one oddity though. The flight engineer said that, despite a restless night at the guest house, he slept about three hours in Tallinn before the flight and he felt fresh. The report also quotes him as saying that for no specific reason, he felt thoughtful during the accident flight and could not concentrate. That doesn’t sound like “fresh” to me.
That evening at Tallinn Airport, the first officer was the Pilot Flying.
After uneventful flight preparations, the flight crew taxied to runway 08, entering at the west end of the runway. They lined up for take-off. At 16:47:22 UTC, the aircraft began its take-off roll.
In aviation, we use V-speeds to refer to specific speeds that are critical to flight. These speeds vary from aircraft to aircraft and even flight to flight, so it is important to know what the monitoring flight crew member is referring to when calling out the speed reached on the ground or in the sky. There are two V-speeds relevant to this incident.
V1 is the speed beyond which the take-off should no longer be aborted.
VR is the rotation speed: the speed at which the pilot gently pitches the aircraft up, which causes it to lift off from the runway.
So on a normal take-off, the monitoring flight crew member, in this case the navigator, would call out when the aircraft reached V1 which means that the speed at which they can safely abort the take-off has been exceeded. On this flight, V1 was 182 kilometres per hour (98 knots, 113 mph).
Then the navigator would call out that the aircraft has reached VR, which is when the Pilot Flying should initiate rotation because the aircraft is now going fast enough to take off. The Pilot Flying should not attempt to lift off before this speed is reached. On this flight, VR was 201 kilometres per hour (108 knots, 125 mph).
So where were we? At 16:47:22, the aircraft began its take-off roll.
Ten seconds later, the Pilot Flying initiated rotation. He pulled back on the elevator controls to lift the nose-wheel off the ground, so that the aircraft would lift up from the runway and take off (rotation). The elevator, which controls the pitch, was positioned to 17º and then released to 9.2º. The aircraft pitch angle increased to 4.6º.
That’s when the navigator made the V1 call.
So this is already a mess. The Pilot Flying started rotation long before the aircraft had reached the right speed. But even worse, the V1 call was early. At that point, the aircraft had reached a speed of 160.5 kilometres per hour (86 knots, 100 mph) which means the take-off could still safely be aborted. The calculated V1 was quite a bit faster: 182 kilometres per hour (98 knots, 113 mph).
When interviewed later, the Pilot Flying couldn’t remember if he’d heard the VR callout before rotation. He said that he rotated the aircraft “at VR or maybe a bit earlier”. Analysis of the Flight Data Recorder showed that the Pilot Flying started rotation at 144.5 km/h. Just to refresh your memory, the VR speed was 201 km/h. They weren’t even close.
It was at this point that the flight engineer called out Retracting! in Polish and selected the lever to bring the gear up.
The landing gear control is located in the cockpit central console. The landing gear on the An-26B should not be retracted until the aircraft has at least fifty metres (164 feet) in altitude and an indicated airspeed of 230 kilometres per hour.
It’s operated by the flight engineer who retracts the landing gear by turning the switch counter-clockwise from the neutral position. The flight engineer does this when commanded to do so by the Pilot Flying.
The flight engineer later said that he saw the Pilot Flying rotating the aircraft and clearly heard him command “Gear Up!”
None of the other crew members heard the command. There is no record of the command on the cockpit voice recorder. The Pilot Flying was clear that there had been no call-out for “Positive Climb”, which would have preceded his command to raise the gear. On the other hand, neither had there been any V-speed call-outs which should have preceded his attempt to take off.
Under normal circumstances, flight crew are prevented from retracting the landing gear on the ground.
When the landing gear is supporting the weight of the aircraft, the landing gear can’t retract: a Weight-on-Wheels switch blocks the electrical current so that the electro-hydraulic retraction system cannot operate. However, the lift on the wings during the end of the take-off roll can sometimes be enough to allow the landing gear to be retracted.
Although the aircraft was moving slowly and at this stage was not actually airborne, when the Pilot Flying pitched up, he would have increased the lift on the wings. It seems that this extended the landing gear struts to the point where the current was passed to the retraction system.
The wheels retracted.
The control is not within the pilots’ field of view. Presuming both the captain and the first officer were focused on looking forward and scanning the instruments, they would not have seen the flight engineer’s action.
To be honest, by this point so much had gone wrong, I just can’t really imagine what they could have done to save the situation even if they had known.
Let’s look at all the calls that didn’t happen.
V1, the speed at which it is no longer safe to abort the take-off, was called out by the navigator, but the call was made after the aircraft had rotated. The call was wrong however, as the aircraft had not reached that speed.
VR, the correct speed at which to rotate the aircraft, was never reached and never called.
Positive climb! (or in the US Positive rate!) is called when the aircraft is climbing away from the airfield at a safe velocity. Only after this call would the command to bring the gear up to be given.
Gear up! is called by the Pilot Flying as an instruction to retract the landing gear. Changes in configuration of the aircraft are generally never made without a direct instruction by the Pilot Flying.
Only one call-out relevant to the take-off was made, which was that the speed at which the aircraft take-off run could safely been aborted had been exceeded. And sadly, even that call was wrong.
As a sideline, there was a great thread on Airliners.net which asked what the opposite call was to Positive climb!. What would the call be if, after rotation, there isn’t a positive rate of climb. The following are my favourite answers:
- “Roll the crash trucks.”
Those words get the famous [expletive] on the accident report when they transcribe the CVR
Technically, the opposite would be ‘negative descent, gear down’. Which would be odd because two negatives make a positive, so the plane is climbing but with its gear down.
“BRACE BRACE BRACE!”
The point is clear: if you have rotated correctly, then a positive rate of climb can be assumed; so much so that, in simulator training, flight crew being tested on their monitoring are caught out calling positive climb without actually checking first. Only after the positive climb call should the gear up call be made. And I can’t believe I’m even having to say this, but only after the gear up call has been made should the gear be brought up.
The accident report makes a similar seemingly banal statement on this.
Rotating at the correct VR will result in aircraft being able to take off and retraction of the gear will not result in runway contact.
Well, that’s the idea anyway…
So the aircraft was travelling along the runway, not having quite hit V1 but still travelling at a fair speed, when the wheels disappeared out from under it.
The aircraft obviously then slammed onto the ground to skid along its belly, although the report charmingly refers to it as “contacted the runway”, as if it were an old Facebook friend.
The blades of the right-hand propeller scraped along the runway. A few metres down, the left-hand propeller left similar marks on its side of the runway. The aircraft slid for 1,228 metres (4,000 feet) before coming to a rest.
The aircraft’s belly was scraped and abrased. The tips of the propeller blades were bent backwards. The left bay door had bent to the left and partly broken off.
As crash landings go for a plane that hadn’t taken off yet, it was a pretty good one.
The core issue here is that the crew did not appear to be working together in any sensible way. The Pilot Flying initiated rotation without waiting for the call from the navigator. The navigator made the initial call early, which meant that he’d stated that the aircraft was travelling too fast for the take-off run to be aborted, whereas it was not too late. Based on the accident report, the captain appears to have done nothing at all. And the flight engineer, presumably triggered by seeing the Pilot Flying pull back on the controls, retracted the landing gear while the aircraft was still (barely) on the ground.
The crew had only worked together for nineteen hours, so they hadn’t had the time to become a team. The Pilot Flying had less than 500 hours experience on type and had trained on a different aircraft type. The flight engineer had only just joined the crew and had not flown for three months before taking the position. Although the records show that all of the crew had gone through CRM training within the past few months, clearly it wasn’t enough.
The accident report cites the flight engineer as the cause of the accident. There’s no question that it was his action which directly caused the aircraft to “contact” the runway. However, the instigating action was the Pilot Flying who initiated rotation without waiting for the VR call. The report also correctly calls out the CRM of the crew and the switch that allowed the gear retraction to be activated in the contributing factors.
The report concludes with a recommendation that the operator “set up and implement procedure with clear verbal exchange between pilot flying and flight engineer confirming the gear operations”. I find this odd, because this procedure already exists and the statements by the crew make it clear that they were aware of the call-out/response sequence, they just didn’t follow it. The second recommendation is more sensible, in that the operator should monitor and supervise their crew better to ensure that they are following procedures. And finally, they recommend that the operator look into the operation of the Weight-on-Wheels switch.
The full accident report can be read in English here: ACCIDENT INVESTIGATION FINAL REPORT AN-26B, SP-FDP Tallinn, Estonia 25. August 2010
- Note: The Estonian Safety Investigation Bureau does not have a link to the Estonian version of this report, if there is one. The English report analysis refers to the VR as 182 km/h which is actually V1. This is clearly an oversight as the speeds are correctly defined in the body of the report. However, there is also a conflict for the speed at which the pilot started rotation, which is cited as 123 km/h in one part of the report and 144.5 km/h in another. As I was not sure which figure was correct, I chose to give the Pilot Flying the benefit of the doubt and use the higher speed, although it is still embarrassingly less than the VR.
I am enjoying an extended stay in Tallinn and it occurred to me that I’d never written about an Estonian aircraft incident…or even read an Estonian accident report. So while I am here, I’d like to make up for this omission and share a collection of accidents and incidents with you. I hope you find the collection as interesting as I do.
The actions and reactions of this crew are beyond comprehension. Has there ever been an investigation into what was carried? Toxic fumes that caused (partial) crew incapacitation? From what I read here, no crew member behaved very rationally.
A late “V 1” call-out should not affect any take-off unless an engine failed. Sylvia correctly mentions that the crew will consult tables that will give them the correct speeds for the runway used, the aircraft weight and configuration and wind and temperatures. I am not familiar with the Antonov and the methods whereby the criteria are calculated but most crews will have pre-computed tables (RTOW or TL tables) that take into consideration all factors including obstacles and criteria in the event of a return with a failed engine, missed approach etc. Sounds a bit complicated and it is. If the crew were to do it for every take-off they would be out of duty hours very quickly without doing much flying. That is why these tables, with all known parameters, are available. They usually come from specialist companies like Flygprestanda in Sweden. Often, if the runway is long enough and criteria are met for a “balanced field” V1 and VR can be the same speeds. My guess is that in this event they could have been identical which is why the English report is not making a distinction.
But: the crew acted in a very strange manner. The captain did not act when he should have noticed a premature attempt to rotate, speed call-outs were faulty, the navigator did not perform coherently nor did the engineer. Does the report mention any further physical examination of the crew? It would have been interesting to find out if any traces of a toxic substance (other than alcohol) were found in their blood.
There are some very odd omissions in the accident report – for example it was specified that there was no alcohol found but it isn’t clear to me whether other tests were done. The engineer is specifically quoted as making his call of retracting in Polish, but at no point is there any discussion of the language used in the cockpit or even the nationalities of the crew.
PS sorry for yet another addition.
For those who are interested, the take-off requirement calculations for a commercial airliner are rather complicated since the aircraft must be able to, at a pre-calculated speed (called V1, the engine failure recognition speed) to either bring the aircraft to a halt or continue the take-off.
But that is only the beginning: The stopping distance may include a stopway, an extension of the runway able to support an aircraft without serious structural damage. On the other hand, if the aircraft continues it must be able to reach a height of 35 ft. above the runway or a clearway, being an area clear of obstacles and under the control of the airport.
The concept of “balanced field” is in fact purely theoretical and not of great value to the crew. It simply means that, if an engine fails at V1 the aircraft must be able to come to a complete stop at the end of the runway (not stopway) OR continue take-off with a failed engine and be able to reach an altitude of 35 feet at the end of the runway (not clearway).. V1 is calculated so that it must be possible to stop OR accelerate with a failed engine to VR and meet the climb requirements. So it is obvious that there are complicated calculations needed to determine V1 and VR. But there are other considerations such as V2 or (popularly) optimum climb speed with a failed engine in order to clear obstacles, clean up and and accelerate to VFTO (final take-off speed). There are calculations referring to obstacle clearance and ability to go-around if the aircraft returns for landing and the crew make a mess of it.
In the case of the Antonov incident it would seem that there had been sufficient runway to bring the aircraft to a safe halt at VR, in which case the two are of course the same.
Even a lay person will understand that the calculations are complex. They would require a thick manual, the AFM (aircraft flight manual), with lots of charts and the AIP (a government publication) which among other information gives all details about runway, elevation, obstacles, etc. The TL tables have most of the calculations done for the crew, pertaining to a particular airport, runway, obstacles and of course the type of aircraft. All the pilots have to enter are variables such as the aircraft take-off weight, air temperature and wind. This will tell them whether take-off is permitted given a certain weight and atmospheric conditions. The tables are given for particular runways and issued by commercial companies that can buy them.
We once spent a whole day preparing the tables for a Fokker F27 manually as Flygprestanda could not supply them for the airport we were intending to use (Kassel in Germany). Fortunately, there was only one runway.
Last but not least: it has been mentioned that the crew had undergone recent CRM training. The fact that the crew had not previously worked together should not have been a factor, It has happened that I flew with a copilot who had been with the company for a few years on the same type of aircraft, yet I had never even met him before. CRM and SOPs (Standard Operating Procedures) are there to ensure that every crew member is trained to do exactly what is expected, regardless of how often they have worked together.
What I failed to mention: In this particular incident the “V” speeds had become irrelevant and invalid.These speeds, especially “V-1” are calculated so that sufficient runway or stopway will remain if the take-off is abandoned. Premature rotation would have caused drag and prevented the aircraft from getting airborne at the pre-calculated point, compromising the margin of room before the end of the runway / stopway.
On some (very) long runways V1 may equal Vr: even just at the point where the aircraft starts its transition from accelerating on the ground to getting airborne, the remaining runway may allow the crew to stop even at such a late moment.
Rotating before the aircraft can fly changes everything. The “V1” speed will be reached, depending on the pitch angle, at a much farther point than where it should have been as calculated. The “Rotate” call would have had meaning as this refers to an aerodynamic speed..